Slashdot Mirror
No More Rebooting?
blankmange writes: "This headline caught my eye: 'The End of Computer Rebooting.' Seems that there has been some new developments in memory technology: The new thin-film technology that could give rebooting the boot is based on resistor logic rather than the traditional transistor logic used in most PCs and other memory-enabled devices. It also is considerably faster than current memory systems and holds the promise of reducing the time required to transfer and download multimedia content and other massive files. This is great news, but what am I going to do with the extra hour or so a day?"
True, but suspend still drains power from a battery, and hibernate uses hard disk space. Besides, even from hibernate, you need time to read data from Hard disk, which does take some time.
This solution will mean no power consumption and no data loss. Plus, heat inside a casing will greately reduce. Plus computers can get smaller and big bulky hard disks vanish.
My mom never taught me to sign.
I did, and it's about 30 dollars a year, assuming the worst case, that your computer is consuming 300 full watts all the time, which it isn't.
I've had enough abrasive sigs. Kittens are cute and fuzzy.
People: Read the frelling article. This isn't 'an end to rebooting', it's highspeed nonvolatile memory that could theoretically be used to replace mass storage and RAM simultaneously. Although this would speed up booting a bit, it would not obviate rebooting entirely.
In fact, on some OSen (cough, Windows, cough) it could be very dangerous - if there's only one copy of the OS code in this combination memory, you can't reboot and reload a fresh copy from disk - meaning bugs have a significantly greater probability of rendering your system unusable.
Sounds like fun, right?
--
Damn the Emperor!
This is what memory was like in the olden days, before DRAM took over from core memory. Apparently early DRAM computers had big piles of batteries in the bottom of the RAM cabinets to prevent the memory from failing due to a power cut.
(We're talking early 70s here; before my time)
Transistors ain't analog? You go down to a low enough level and everything's analog. The binary part just means there are two states we can map '1' and '0' to: two voltages, two current levels, two different resistor values, two different amounts of charge, etc. These magnatite thingies can probably be put in (at least) two states with different resistances. To sense it, they run a little bit of current through it and measure the voltage on the output. Or something like that.
The article is not even clear on whether this development is supposed to be replacing RAM or hard disks. But either way, it cannot eliminate the need for rebooting. The primary reasons for rebooting are either to reset the operating system to a known state, or to upgrade low-level software (such as the kernel in Linux, or your web browser in Windows). Neither of these necessities go away with non-volatile RAM, regardless of how fast, cheap, or capacious it may be. These are software issues, and they need software solutions.
use constant PERL_IS_BROKEN => $] >= 5.006;
Don't you know that every speed improvement will speed up your internet access? Like upgrading your processor to a Pentium 4 will turn that modem into a T1!!! It's just buzzwords! The only speedup will be that the servers now have more memory bandwidth and can handle more connections at once.
The whole article is mistitled. It won't be an end to rebooting, it will be an end to cold booting.
If you want to eliminate the reboots from Windows, tell bill gates and co to make it more modular and less inter-dependent so you can insert and remove drivers just like *nix kernel modules.
I found that Watt numbers on the power supply are quite a bit more than the real consumption. The number is the maximum allowed, and is reached only during start-up, if ever.
To make an example, I admin a small Linux cluster with 6 PCs. Each of them is a 1.3 Ghz Thunderbird, with 1.5 gigs of PC133 RAM, an average HD, and a 300W power supply. Most of the time the processors are at 100% load.
All 6 machines are powered with a 2000 VA UPS. An UPS like that is capable of delivering around 1600 Watts, but the load indicator never surpassed the mid level. So I would say that 150 Watts are more than enough for an average PC to operate. The 300-400 watts are for booting the things and spinning the drives up.
This is an appalling summary - and the article is no better.
"The technology is highly suitable for broadband Internet connections, Hsu said, noting that it combines the features of low voltage, high speed and low power consumption."
Yes, fantastic. That's great for those broadband internet connections. Faster memory is always good, but choosing this as an application is just a moronic use of buzz words.
"Ignatiev said the new technology is about 1,000 times faster than flash, which is nonmechanical and currently the speediest memory on the market. "
Flash memory is the fastest type of memory on the market? No, it is a form of non-volatile memory, which is very slow by RAM standards.
"is based on resistor logic rather than the traditional transistor logic"
Actually, you'll find that DRAM in most modern computers are capacitative devices - the techniques to make them are the same as MOS transistors, but they do not use switching to store values, IIRC.
I wish people would not spout such rubbish.
Interesting highlights:
The trasentric paper quoted Electronic Buyer's News:
The interesting elements of this:- Much of this research is funded by a DARPA contract which means it is the money of US Taxpayers at work.
- Samsung is part of the same contract.
Methinks that perhaps Samsung and IBM are using the same (or very similar) technology.The Wired article is fairly lengthy and also details the biography of Stuart Parkin. Parkin is the IBM fellow that has been driving most of the MRAM research.
Ciao.
The reason for the original poster picking on Windows is that Windows, in most folk's experience, has to be rebooted frequently because of error accumulation. In other words, I could leave my Win98 box running (with no additional applications up) for 12 hours and when I came back it would be locked up. This isn't about shutting down during upgrades or installs, it's about shutting down because of frequent OS corruption during everyday use. In this case, you *need* memory to clear itself out.
"Can I say you're my lovepuppy?" Founding member of SODAMNHOTT
Just remember - the uptime counter rolls at ~497 days. :-)
Scared the Hell out of me when I saw my DB server with an uptime of 23 days...until I realized that it had indeed been up for 520 days. This machine gets *hammered*, too.